Re: [Help-glpk] Slow performance on "Select minimum" task

[Jan van Rijn]

2018-06-05 23:39 GMT-04:00 Michael Hennebry <address@hidden>:

On Tue, 5 Jun 2018, Jan van Rijn wrote:

I need to add two things:
- It is my conjecture that there should be an additional constraint, i.e.,
y[r,c] > 0 (Otherwise non-selected rows could participate towards a lower
score)

Correct,

- M[r,c] should contain positive values (which guarantees that y[r,c] == 1
iff x[r] - SUM x[s] == 1)

I'm pretty sure that is not necessary.
the y's depend only on the x's and the order of the M values.
In any case, I think the zeros in your original problem
will not be much of an issue.

I should rephrase: The array should contain values >= zero.

(negative values are an issue, but these can be scaled away easily. )

Thanks again!